How I Sharpen – A Quick Overview
Posted on 15 October 2009
I just posted something elsewhere about how I sharpen for prints and I figured I’d get some extra mileage out of it by posting it here as well. A few disclaimers first, though:
- “Quick?” OK, the post is long. But there are whole books on sharpening, so by contrast I think this qualifies as a quick description. In fact, I’ve left a lot out of the description!
- There are people with far more expertise on this topic than I have and I have no illusions that this description represents the “right” way to do this, much less the “best” way!
The subject of how to sharpen photographs in post for print or electronic output is one that confuses many people… and a subject to which many books, online article, and forums posts have been devoted. There are any number of ways to get the desired results via sharpening, and different techniques are called for depending upon taste, the nature of the image, and the final form of presentation – size? print? jpg? etc…
Here is a very general description of what I do when I print. I’ve left some variations out of this description. The description also covers software that I use in my workflow – you might prefer something different, but you might still be able to adapt these ideas. You’ll note at least one controversial method later in the list, but try it before you dismiss it. The approach I use could well be “over-kill” if you just want to pump out a bunch of jpgs to share with friends and family or if you want to make some small prints – my end goal is good sized prints, and I work on each one rather carefully rather than mass-processing them and printing a bunch at one time.
And please understand that I’m most certainly not implying that my way is the right way. It works for me, and that people who view my prints often remark on their sharpness and detail. (And a few other things, too, I hope! :-)
1. Shoot RAW and, of course, pay careful attention to stuff that affects image sharpness – choice of lens, aperture, very careful focus, consideration of DOF issues, consideration of diffraction blur issues, tripod, mlu or live view, remote release, etc.
2. Using ACR (Adobe Camera Raw) to do my conversion I remove any dust spots, try to maximize image dynamic range without losing shadows or blowing highlights, perhaps do some color correction, etc.
3. I may do some sharpening in ACR or I may not. ACR sharpening is great for images in which you might have a bit of noise that you do not want to sharpen but in which you also have some distinct edges that might benefit from a little input sharpening. This is especially true if your image includes some sharp edges along with other areas of uniform color/luminosity or smooth gradients. You can use the masking feature in ACR’s sharpening window to keep the sharpening away from the gradients/uniform areas.
4. I open the image as a smart object in CS4 in 16-bit mode.
5. Although many of the post-processing techniques I may apply in CS4 are not strictly “sharpening” techniques, it is important to note that the subjective impression of sharpness in a photograph is the result of many things other than strictly sharpening techniques. For example, I may do a lot of work with masked curves to increase contrast in certain areas of the image, to recover shadow details and so forth. You can do all of the “sharpening” in the world, and your print may still look dull if you don’t deal with this stuff.
6. On to actual sharpening…
7. Using a smart filter (e.g. – selecting the bottom smart layer, and then selecting from the Filter… menu) I first apply smart filtering. Actual settings can vary depending upon the nature of the image, but a decent starting point for my 21MP FF images might be 150, 1.0 – I may well lower the second value a bit if I notice halos near edges, since doing so makes them narrower. On images with a lot of very fine detail I might begin with an alternative setting of 300, .3. (The latter is based on an old recommendation from Canon.) I watch the screen at 100% magnification while I do this – I want to verify the quality of the sharpening and make sure that there won’t be visible “sharpening halos” around high contrast edges – if there are, I reduce the second value first, and in a few cases I may need to reduce the first value as well. (Or I may have to resort to trickier approaches such as varying the amount of sharpening in different areas of the image – but I’m not going to try to explain that here.)
8. Still working on the smart object layer I now add an Unsharp Mask (USM) filter as well. Typical starting point settings might well be on the order of 12, 50, 1. Again, the values (especially the first one) may vary depending upon the image. By the way, I reduce the display magnification for this operation so that I can see the entire image on the screen at once, and I watch carefully to avoid wide halos around edges where uniform light and dark areas meet.
9. I save the entire CS4 file will all layers intact. I duplicate the image and flatten it, and from here on work on the flattened copy rather than the saved original. What I do next depends upon whether I’m printing or making a small jpg for online presentation.
10. If I am printing I use the image size dialog to set the print size in inches that I’ll need. I DO NOT UPREZ. In other words I let the “resolution” end up being whatever it ends up being when I select my target print size.
(At this point a few of you are gasping “What the hell!?” You have “learned” that cumbersome and expensive uprezzing processes are required to get good large prints. I used to think so also until I heard Jeff Schewe explain this approach. I was skeptical, too, but I tried it, and discovered that it really works very well in most cases. Basically your printer has built-in software that does interpolation, and there are arguments against interpolating first in CS4 and then interpolating a second time in the printer. On a very small print the resolution might be 500+ and on a very large print it might get close to 180 – but this actually makes a lot of sense if you think about it a bit. Note that if your resolution drops below about 180 in a large print, all bets are off and other approaches are likely needed.)
11. Now, at 100% magnification, I do another smart sharpen operation. The idea here is to over-sharpen a bit – at 100% on the screen the image should look a bit “grainy” or “crunchy” – not pretty at all. The goal is to compensate for the ink spread (also known as “dot gain”) that will take place when this slightly over-sharpened image ends up on the paper. You have to do some testing to figure out the best values, and they will be different for different types and brands of paper. I print on Ilford Galerie GFS for the most part, where typical “output sharpening” settings might be on the order of 200, .3. The first number might to as high as perhaps 240 or as low as perhaps 100, and the second value might be .4 in some cases.
(If I’m going to jpg, I obviously have to down-rez the image to smaller size. Without going into the gory details here, a final USM process is useful here after downsizing. Try 35, 1,1. Some like the first number a bit higher, but I’m more likely to go a bit lower in some cases. UPDATE: I’ve recently tried the method for web sharpening described by Ian Plant and I think it may produce better results than that approach I’ve been using.)
(Credit Where Credit is Due Department: Charlie Cramer helped me understand the basic concept of the “output sharpening” for printing and showed me some starting points for determining settings. The approach of not uprezzing and the logic behind it is explained in the very useful “From Camera to Print” video from Luminous Landscape featuring Jeff Schewe and Michael Reichmann. And I apologize in advance for any errors I’ve made in my adaptations of their systems. :-)